[Qemu-devel] direct usage of rt_clock instead of rtc_clock

[Qemu-devel] direct usage of rt_clock instead of rtc_clock

[Paolo Bonzini]

Hi Peter and Andrzej,

the following four files in hw/ are using rt_clock directly

   omap1.c
   pxa2xx.c
   strongarm.c
   twl92230.c

instead of rtc_clock.  This means that they're not completely 
deterministic when using -icount or in the future qtest.  A simple 
search-and-replace would be okay I think.  However they could also use a 
notifier to update the timers whenever the rtc_clock jumped backwards 
(see commit 17604da).  If you are going to add the notifier I'll happily 
leave the work to you; otherwise I'll post the search-and-replace patch.

Paolo

Re: [Qemu-devel] direct usage of rt_clock instead of rtc_clock

[Peter Maydell]

On 18 January 2012 18:04, Paolo Bonzini <pbonzini@redhat.com> wrote:
> the following four files in hw/ are using rt_clock directly
>
>  omap1.c
>  pxa2xx.c
>  strongarm.c
>  twl92230.c
>
> instead of rtc_clock.  This means that they're not completely deterministic
> when using -icount or in the future qtest.  A simple search-and-replace
> would be okay I think.  However they could also use a notifier to update the
> timers whenever the rtc_clock jumped backwards (see commit 17604da).  If you
> are going to add the notifier I'll happily leave the work to you; otherwise
> I'll post the search-and-replace patch.

I have no strong interest in any of these platforms -- if the simple
search-n-replace you propose works OK and has no particular disadvantages
over the code as it stands that's fine with me.

-- PMM